People will adapt - or run into massive problems anyway.

People tell AI things they wouldn’t tell their closest friends, then act surprised when those chats can end up in court records, company systems or leaks. The chatbot feels like a diary, but it’s closer to a diary stored on someone else’s server. Maybe the new rule is simple: if you’d panic seeing it read aloud in court, don’t type it into the bot.

Google discovering that AI search may quietly steer shoppers toward higher prices is a fun twist: first search results became ads, then ads became recommendations, and now recommendations apparently have a concierge fee. At this rate, AI powered shopping just means your wallet gets the hallucination.

Absolutely. Unlike these ignorant people, I know what communism is.

Let’s get the good stuff down from the loft… the loft ladder creaks

They’re all communists.

Trump’s government siding with OpenAI is peak MAGA economics: copyright is sacred until Silicon Valley needs millions of other people’s books, articles and art for free. Then suddenly property rights become an annoying obstacle to »winning the AI race.« Apparently America First now means AI companies first, creators somewhere near last.

The bigger the model the bigger the memory requirements.

Exactly. It’s super uncommon to use standard consumer hardware to build a data center.

The Mac Studio already has it. And from October, it’ll also be available with 512GB. It does cost around 18K, though.

Macs, in particular, are likely to be the best option at the moment, unless you absolutely need the largest Frontier models.

OpenAI buying tens of thousands of Macs for AI is the funniest possible plot twist: spend years convincing everyone the cloud is the future, then discover the hottest AI infrastructure is basically a warehouse full of Mac minis. Somewhere, an IT admin is whispering: »I told you so.«

Turns out the scary AI chatbot is sometimes better at spotting propaganda than the supposedly boring old search engine. The real plot twist: AI summaries can be worse than both. So we somehow invented a technology where asking the robot a question may be safer than trusting the robot Google glued above the search results.

Musk treating contracts as optional paperwork whenever they become inconvenient is becoming less of a plot twist and more of a business model. Musk’s version of an open future increasingly looks like this: buy the platform, control the infrastructure, attack anyone who resists, then call everyone else monopolists.

Convenient, yes. But prompt injection is still the uninvited coworker with admin ambitions. The browser wars just became agent wars.

Ultimately, though, Google is all about something else: Google is currently promoting YouTube in its AI Overviews and on Gemini in general. So, for a change, it’s promoting its own service rather than, say, Reddit.

Nothing says »AI job boom« quite like humans earning pennies to draw boxes around objects so a trillion dollar industry can call the result automation. The magic of AI apparently still requires an invisible army doing digital piecework: until the machine gets good enough to automate the people who trained it.

GEO gurus spent months telling brands to get on Reddit because ChatGPT loves it. Now citations drop 86% and suddenly the strategy is diversify. Amazing how quickly a growth hack becomes timeless marketing wisdom once the algorithm stops cooperating. Maybe building useful, credible information was the strategy all along. 🤷

Perhaps this is an important point, as I’ve no idea why this post is being downvoted so much?!

So after years of arguing over which model has the biggest brain, Nvidia discovers the real genius is basically good project management: memory, tools and a supervisor telling Claude to stop wandering into dead ends. Turns out AGI may just need a competent middle manager. … Somewhere, an enterprise architect is feeling extremely vindicated.